應用電腦視覺技術於PCB自動檢測系統之設計及開發

Abstract

於印刷電路板組裝業中，所應用之表面黏著技術(SMT)日益成熟，著裝機之準確度亦不斷提昇，但仍有許多瑕疵存在於製程之中，其中主要包括元件之缺件、歪斜、極性反、錫膏造成短路等現象。 由於現場生產實務問題，本研究提出以虛擬CCD對所開發之電腦視覺檢測系統模組化為離線訓練作業系統及線上檢測作業系統，並使其具有自動定位之功能。並針對(1)片狀電容之缺件、歪斜與(2)具IC腳之排阻、小型外引腳積體電路(SOP)及方形扁平封裝積體電路(QFP)之橋接等瑕疵現象，進行檢測演算法之開發。 鑑於不同照明方式會密切影響檢測結果，本研究採用多層式環狀發光二極體陣列(LED Array)，可根據不同取像需求，以程式控制自動切換照明方式，取得適當之影像，因此本研究利用所開發之演算法，搭配不同照明以決定最適光源與PCB自動檢測演算法之組合，並以實作系統來驗證所開發PCB自動檢測系統之適用性及有效性。

SMT machine in PCB assembly is moving up rapidly during the past decades. While many defects, such as component missing, misalignment, wrong polarity, solder bridge, etc., still hide in the working process. This paper is to design and develop an SMD-based PCB auto-inspection system by using the machine vision technology. The in-line PCB production problem is considered. The mainly focused inspection items of SMD components are capacitor missing and misalignment, and solder bridge of the lead of SOP or QFP. A “virtual CCD” concept is proposed to aid the modeling of the auto-inspection system as an off-line training system and an on-line inspection system. This also makes the developed auto-inspection system have better capability in auto-capturing the PCB position in production line. That is, the PCB to be inspected is allowed to have a large degree of positioning freedom in production line. A new programmable-controlled lighting environment composed of multi-layer of ring-type LED array is also designed, which can provide appropriate light for grabbing the desired PCB image. A prototype of the PCB auto-inspection machine is implemented which shows high robustness and effectiveness of the proposed machine vision method.